請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/44508
標題: | 以統計觀點為主之阿茲海默症疫苗的成本效益分析 Statistical Aspects of Cost-Effectiveness Analysis of Vaccination for Alzheimer’s Disease |
作者: | Kuen-Cheh Yang 楊昆澈 |
指導教授: | 陳秀熙(Tony Hsiu-Hsi Chen) |
關鍵字: | 設限資料,成本效益分析,阿茲海默症,免疫治療, Censored data,cost effectiveness analysis,immunotherapy, |
出版年 : | 2011 |
學位: | 碩士 |
摘要: | 研究背景:阿茲海默症是一個退化性的慢性疾病,也是最常見的失智症原因,雖然有藥物可延緩疾病的退化,但目前並沒有根治的方法。自1990年代以來,阿茲海默症的免疫治療,陸續進入不同期的臨床試驗,但目前還尚未有成本效益方面的研究。然有關醫療成本效益常面臨疾病的治療過程,病人中斷治療,而出現設限狀態,但中斷治療後照護成本和疾病演化仍是繼續的情況下,若忽略此設限狀態的存在,將導致成本與效益的估計偏差。而且,因涉及疾病自然病史狀態之間的轉移,須利用較複雜馬可夫決策模式。此外,成本與效益之間的相關性亦常被忽略,本論文以阿茲海默症免疫治療成本效益分析為例,示範如何針對這些問題進行處理。
研究方法:本研究以文獻中估計所得的疾病自然史的轉移機率,以馬可夫決策模型模擬阿茲海默症的族群作為研究世代。比較的策略則分為為施打阿茲海默症疫苗的介入組與未施打的控制組,每組各為1000人,共追蹤6年,並得到此在未有設限狀態下之機率性成本效益分析之結果。繼而模擬存在某一設限機率下的資料,並以Direct (Lin)法與逆加權估計量法(inverse probability weighting, IPW)去校正設限狀態下的成本與效益的估計,效益面指標則分別以未考慮生活品質之存活人年及考慮生活品質之品質校正後存活人年,分別進行機率性成本效益分析,以付費意願(Willingness to pay) 10,000美元為閾值的情況下,看其成本效益接受曲線(cost-effectiveness acceptability curve, CEAC)去評估介入組相對於對照組具有成本效益的機率。 結果:(1)馬可夫決策模型的模擬結果,以品質校正後存活人年為效益指標時,介入組相對於控制組之ICER為$3000/品質人年(95%CI: $-12,000~$14,000);增加淨效益(incremental net benefit; INB)為$2,729 (95% CI: -$1,003~$1,1634);介入組具成本效益的機率為88.4%。(2)若考慮存在設限情況下,以Direct (Lin)與IPW所估計的成本,都比未考慮設限資料存在或沒有設限資料存在的直接估計值來的高。存活人年的增加成本效益比(incremental cost-effectiveness ratio; ICER)估計,以Direct (Lin)和IPW分別為$9,310/人年 (90% CI: $4402~$13,322)和$6,987/人年(90% CI: $937~ $11,646);品質校正後存活人年的ICER,以Direct (Lin)和IPW估計,分別為$12,885/品質人年 (90% CI: $5,808~$19,132)和$8,955/品質人年(90% CI: $1,066~$1,6319); 以IPW估計下,效益指標以存活人年和品質校正後存活人年來看時,介入組具有成本效益的機率分別為85%和59%; 若以Direct (Lin)估計下,以存活人年為效益指標時,介入組具有成本效益的機率為60.5%,以平均品質校正存活人年為效益指標時則為24%。 結論:透過阿茲海默症疫苗的成本效益分析,本論文示範如何利用馬可夫決策模式配合Direct (Lin)與IPW考慮設限資料下,如何解決未考慮設限資料存在而直接估計時,所造成的低估現象。 Background: Alzheimer’s disease (AD) is a degenerative chronic disease, also the most common form of dementia. Although some current medications may delay the progression, it is not possible cure for AD. In 1990s, studies on immunotherapy for AD have been published. Since then, more immunotherapies with a clinical trial design entered the different stages. However, no economic valuation for the cost-effectiveness of immunotherapy for AD was performed. Vaccination against AD is illustrated to this study. However, the censored data are a common feature in clinical trials. If we ignore the censored problem, bias of estimation would occur because the cost and efficacy still accumulate after censoring. Additionally, the complicated Markov model would be applied to different status of disease nature course. Finally, the relationships between costs and effectiveness are often ignored. Therefore, this thesis used the cost-effectiveness analysis of vaccination for Alzheimer’s disease as an example to illustrate the methods how to resolve above problems. Methods: We used a Markov cost-effectiveness model to construct the nature course of AD. The micro-simulation was used to create a hypothetical cohort. The transition probabilities were extracted from previous Taiwanese studies. The treatment group is the participants with the uptake of vaccination and the control group is those without vaccination. The duration of follow-up is 6 years and each group consists of 1000 participants. First, the cost and effectiveness were measured without considering censoring data. Furthermore, they were measured by Direct (Lin) and Inverse probability weighting (IPW) to make allowance for censoring data. The outcome of interests included person years and quality-adjusted life year (QALY). Given the threshold of $10,000 of willingness-to-pay (WTP), we evaluate the probability of being cost-effectiveness for the treatment group by cost-effectiveness acceptability curve (CEAC). Results: (1) Without considering censored data, probabilistic analysis showed ICER was $3000 per QALY (95% CI: $-12,000~$14,000); INB was $2,729 (95% CI: $-1,003~11,634). Given the threshold of $10,000 of WTP, the probability of being cost-effective for the treatment group versus the control was 88.4% in terms of QALY. (2) Considering censored data, the costs estimated by Direct (Lin) or IPW were higher than those without considering censored data. The ICER of mean survival time was $9,310 per person-year (90% CI: $4,402~$13,322) estimated by Direct (Lin) and $6,987 per person year (90% CI: $937~$11,646) estimated by IPW, respectively. The ICER of mean quality-adjusted survival time was $12,885 per QALY (90% CI: $5,808~ $19,132) and $8,955 per QALY (90% CI: $1,066~16319) by Direct (Lin) or IPW, respectively. The mean survival time and QALY for AD were not cost-effective given the threshold of $10,000 of WTP. Given the threshold of $10,000 of WTP, the probability of being cost-effective for the treatment group were 85% and 59% in terms of mean survival time and QALY, respectively, by IPW estimation. The corresponding figures were 60.5% and 24% in terms of mean survival time and QALY, respectively, by Direct (Lin) estimation. Conclusions: By using Direct (Lin), IPW methods and Markov decision model, we demonstrated adjusting for censoring could adjust for censoring lead to downward estimation using an illustration of vaccination against AD. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/44508 |
全文授權: | 有償授權 |
顯示於系所單位: | 流行病學與預防醫學研究所 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-100-1.pdf 目前未授權公開取用 | 1.37 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。